Potential and Uncertainties of RejectClass in Acute Kidney Graft Dysfunction: An Independent Validation Study

Abstract

Background: Kidney graft rejections are classified based on the Banff classification. The RejectClass algorithm, initially derived from a cohort comprising mostly protocol biopsies, identifies data-driven phenotypes of acute rejection and chronic pathology using Banff lesion scores. It also provides composite scores for inflammation activity and chronicity. This study independently evaluates the performance of RejectClass in a cohort consisting entirely of indication biopsies.

Methods: We retrospectively applied RejectClass to 441 patients from the German TRABIO (TRAnsplant BIOpsies) cohort who had received indication biopsies. The primary endpoint was death-censored graft failure during 2 y of follow-up.

Results: The application of RejectClass to our cohort demonstrated moderately comparable phenotypic features with the derivation cohort, and most clusters indicated an elevated risk of graft loss. However, the reproduction of all phenotypes and the associated risks of graft failure, as depicted in the original studies, was not fully accomplished. In contrast, adjusted Cox proportional hazards analyses substantiated that both the inflammation score and the chronicity score are independently associated with graft loss, exhibiting hazard ratios of 1.7 (95% confidence interval, 1.2-2.3; P = 0.002) and 2.2 (95% confidence interval, 1.8-2.6; P < 0.001), respectively, per 0.25-point increment (scale: 0.0-1.0).

Conclusions: The composite inflammation and chronicity scores may already have direct utility in quantitatively assessing the disease stage. Further refinement and validation of RejectClass clusters are necessary to achieve more reliable and accurate phenotyping of rejection.

Graphical abstract

FIGURE 1.

Flow chart of patient selection. Patients were selected from the TRABIO (TRAnsplant BIOpsies) registry to apply the RejectClass algorithm. Complete 1-y follow-up data were available for 441 patients. The required variables for performing the RejectClass algorithm were not available for all patients: because of missing data, acute RejectClass could only be performed in 257 patients, and chronic RejectClass in 422 patients. For acute and chronic RejectClass different strategies were used in the handling of repetitive biopsies, as indicated. DSAs, donor-specific antibodies; PVN, polyomavirus nephritis; w/o, without.

FIGURE 2.

RejectClass phenotypes and the Banff classification. A, This polar plot provides a simple aid to visualize the clustering process and the principal component analysis of the Banff lesions. It illustrates the inflammation score of the biopsies by how far they are from the center, and it also shows a measurable characteristic of the biopsy’s phenotype (represented by the theta angle in degrees). Essentially, this angle is a scaled representation reflecting the multidimensional relationships—or the complex interactions across multiple variables—between the Banff lesions. Thus, instead of relying on black-and-white classification, the theta angle may allow for a more nuanced, fluid interpretation of the phenotype. For each cluster, phenotypic descriptions had been determined in the original report. Cluster 1: no inflammation & DSA^neg (n = 122), cluster 2: DSA^neg MVI (n = 11), cluster 3: TCMR-like (n = 52), cluster 4: no inflammation & DSA^pos (n = 38), cluster 5: AMR-like (n = 7), and cluster 6: MR-like (n = 27). B, Presence of Banff lesion scores within the acute RejectClass clusters. The red line in each bar represents the prevalence of the lesion (score > 0) in the derivation cohort (detailed information on the score distribution of the derivation cohort’s variables is provided in the Supplemental Material [SDC, http://links.lww.com/TP/C958]). Mann-Whitney U test with Bonferroni adjustment for multiple testing was used to compare the Banff lesion score distribution of a variable in this study with its counterpart in the corresponding cluster in the derivation cohort. C, Contingency heatmap showing the association between acute RejectClass clustering and Banff diagnoses. Darker shading indicates a higher proportion of Banff diagnoses in the corresponding RejectClass cluster. D, Polar plot of clustering by the chronic RejectClass algorithm and the chronicity score, analogous to A. For each cluster, a phenotypic description had been determined in the original report. Cluster 1: no chronicity (n = 297), cluster 2: IFTA (n = 24), cluster 3: IFTA & vasculopathy (n = 44), and cluster 4: glomerulopathy & vasculopathy (n = 57). E, Presence of Banff lesion scores within the chronic RejectClass clusters. Derivations in the Banff lesion score distribution compared with the derivation cohort were determined as described in B. F, Contingency heatmap showing the association between chronic RejectClass clustering and Banff categories of chronic pathology. Darker shading indicates a higher proportion of Banff diagnoses in the corresponding RejectClass cluster. *P < 0.05, **P < 0.01, ***P < 0.001. ah, arteriolar hyalinosis; AMR, antibody-mediated rejection; C4d, complement 4d staining; cg, glomerular basement membrane double contours; ci, interstitial fibrosis; ct, tubular atrophy; cv, vascular fibrous intimal thickening; DSA, donor-specific antibody; g, glomerulitis; gs, glomerulosclerosis; i, interstitial inflammation; IFTA, interstitial fibrosis and tubular atrophy; mm, mesangial matrix expansion; MR, mixed rejection; MVI, microvascular inflammation; NR, no rejection; ptc, peritubular capillaritis; t, tubulitis; TCMR, T cell–mediated rejection; thrombi, thrombotic microangiopathy; v, intimal arteritis.

FIGURE 3.

Outcomes of acute RejectClass clusters and Banff groups in the TRABIO (TRAnsplant BIOpsies) registry cohort. A, The risk for death-censored graft failure during 2-y follow-up in the acute RejectClass clusters was analyzed by computing Kaplan–Meier (KM) survival curves (with a log-rank test), and by Cox proportional hazard (PH) regression adjusted for multiple covariates (as indicated). The cluster descriptions as determined in the original report were as follows: cluster 1 (no inflammation & DSA^neg), cluster 2 (DSA^neg MVI), cluster 3 (TCMR-like), cluster 4 (no inflammation & DSA^pos), cluster 5 (AMR-like), and cluster 6 (MR-like). B, The risk for death-censored graft failure in the Banff diagnostic groups was assessed by KM analysis and multivariate Cox PH regression, analogous to A. C, The biopsies were grouped according to the sum of weighted acute graft lesion indicating the inflammation score: “no inflammation” (0.00 to ≤0.04, n = 67); “minimal” (>0.04 to ≤0.10, n = 19); “mild” (>0.10 to ≤0.24, n = 95); “moderate-to-severe” (>0.24 to ≤0.42, n = 52); and “very severe” (>0.42, n = 44). D, The risk for death-censored graft failure associated with the inflammation score was assessed by KM analysis and multivariate Cox PH regression, analogous to A. AMR, antibody-mediated rejection; CCI, Charlson comorbidity index; DSA, donor-specific antibody; eGFR, estimated glomerular filtration rate; HR, hazard ratio; MR, mixed rejection; MVI, microvascular inflammation; NR, no rejection; P, probability value; TCMR, T cell–mediated rejection.

FIGURE 4.

Outcomes of chronic RejectClass clusters and Banff groups in the TRABIO (TRAnsplant BIOpsies) registry cohort. A, The risk for death-censored graft failure during 2-y follow-up in the chronic RejectClass clusters was analyzed by computing Kaplan–Meier (KM) survival curves (with a log-rank test), and by Cox proportional hazard (PH) regression adjusted for multiple covariates (as indicated). The cluster descriptions as determined in the original report were as follows: cluster 1 (no chronicity), cluster 2 (IFTA), cluster 3 (IFTA & vasculopathy), and cluster 4 (glomerulopathy & vasculopathy). B, The risk for death-censored graft failure in the Banff categories of chronic pathology was assessed by KM analysis and multivariate Cox PH regression, analogous to A. C, The biopsies were grouped according to the sum of weighted chronic graft lesion indicating the chronicity score: “minimal” (0.00 to ≤0.30, n = 227); “moderate” (>0.30 to ≤0.50, n = 96); “severe” (>0.50 to ≤0.60, n = 36); “moderate-to-severe” (>0.60, n = 58). D, The risk for death-censored graft failure associated with the chronicity score was assessed by KM analysis and multivariate Cox PH regression, analogous to A. AMR, antibody-mediated rejection; CCI, Charlson comorbidity index; eGFR, estimated glomerular filtration rate; HR, hazard ratio; IFTA, interstitial fibrosis and tubular atrophy; P, probability value; TCMR, T cell–mediated rejection.